The Autonomous Network
In active development. Pieces are rolling out as Diamond facets on the Grid contract. The vision: no team at the center, no single point of failure, no one who can turn it off.
Every AI API you’ve used can be unplugged. Stripe can freeze your account. AWS can suspend your inference endpoint. OpenAI can deprecate a model out from under your product. That’s not a flaw in their operations — it’s the shape of centralized services.
AIPG is being built so that no single entity can stop the inference. Workers register themselves on-chain. Jobs route through cryptographic commitments. Payments settle in smart contracts. Validators police results with their own stakes on the line. The team builds the protocol. The protocol runs itself.
What’s Already On-Chain
These pieces live on Base today as facets of the Grid Diamond contract:
WorkerRegistry— workers bond AIPG and publish their endpoints on-chainJobAnchor— every completed job commits a result hash on-chain, creating an audit trailModelVault— the canonical list of supported models and their reference hashesRecipeVault— image-generation recipes (prompts, samplers, seeds) anchored for reproducibilityStakingVault— non-custodial bonds, no admin keys over user fundsRoleManager— on-chain permissions (no off-chain ACLs)
Look them up in Developers for live addresses.
What’s Coming
Validator-attested job routing
Today the API node picks which worker handles a job. The next step: workers self-claim jobs through a deterministic on-chain mechanism (described in Claim Resolution), and validators attest to whether the work was honest. The routing layer disappears into the contract.
Smart-contract payments
Inference fees flow through the contract instead of an off-chain settlement service. Workers receive payment as soon as a job’s attestation window closes. Validators take a cut. No invoices, no payouts CSV, no team in the middle.
Bonded validators
The big trust upgrade. Validators stake 10K AIPG, sample jobs at random, re-execute on their own hardware, and attest results on-chain. False attestations get slashed. See Validator Nodes.
Dispute resolution
When attestations disagree, the contract runs a fraud-proof protocol — third parties re-execute the disputed job, the loser is slashed, the dispute settles deterministically. No human arbitration.
Why On-Chain Routing Matters
Most “decentralized AI” projects are decentralized in their compute layer and centralized everywhere else. There’s still an API node that decides who runs what, a payment processor in the loop, and a team that can — under pressure — yank the rug.
Putting routing, settlement, and verification into smart contracts changes the shape of the network:
| What | Before (centralized) | After (autonomous) |
|---|---|---|
| Job assignment | API node picks a worker | Workers self-claim via deterministic on-chain rule |
| Payment settlement | Off-chain payment processor | Smart-contract escrow + release |
| Result verification | ”Trust the worker” | Validator sampling + slashing |
| Disputes | Email the team | Fraud-proof in contract, deterministic outcome |
| Failure mode | Provider can turn it off | Network keeps running while workers exist |
This is the same trajectory as Ethereum rollups: start with a centralized sequencer for performance, then progressively decentralize sequencing and proving until the L1 settlement is what gives you trust. AIPG’s “API node” is today’s sequencer. Validators + on-chain routing are how it goes away.
Confidential Computing as the Final Piece
Once jobs route on-chain, payments settle on-chain, and validators verify on-chain, there’s still one open question: who can see the prompts? The host OS of the worker can.
Confidential Computing closes that — prompts and outputs are encrypted in GPU memory, decrypted only inside an NVIDIA-attested TEE on Blackwell hardware. The network can prove that your sensitive workload ran in a secure enclave, and no one — not the operator, not the protocol team — can read your data.
That’s the full picture: autonomous routing, on-chain settlement, validator verification, hardware-isolated privacy. An inference network that nobody can shut down and nobody can spy on.
Read the Deep Design
The full on-chain economics proposal — sampling rates, slashing math, dispute resolution, fee splits — lives at Future On-Chain Architecture. It’s the source-of-truth for what’s being built and how the pieces fit together.
For day-to-day operator and developer docs while the autonomous layer rolls out, see P2P Mode and For Developers.